In breast cancer, the number of lymph node metastases is the strongest predictor of outcome. However, histopathology may underestimate the frequency of metastasis. Here we compare automated molecular detection of cytokeratin 19 mRNA by one-step nucleic acid amplification (OSNA) with histopathology of single tissue sections for the staging of axillary lymph nodes in patients with breast cancer. Axillary lymph nodes were collected from 55 patients with primary breast cancer and sentinel lymph node (SLN) metastases. The central 1-mm portion of each node was processed for hematoxylin-eosin staining, and the remaining tissue was analyzed by OSNA. According to OSNA, histopathology misclassified 41.8% of patients as negative for axillary node metastasis (P=0.007). Of the individual nodes considered negative by histopathology, 4.5% contained micrometastases and 2.5% contained macrometastases according to OSNA. Furthermore, 80% of micrometastases identified by histopathology were reclassified as macrometastases by OSNA. Histopathology failed to identify 81.1% of nodes shown to contain metastasis by OSNA. However, OSNA yielded no false-negative results. On the basis of OSNA results, 3 patients were reclassified to a higher pathologic stage. The number of SLN and non-SLN metastases was unrelated according to OSNA (P=0.891). These results show that, compared with molecular detection, histopathology of single tissue sections significantly underestimates the frequency of axillary node metastases. We discuss the implications of these findings in light of current recommendations on the staging of breast cancer.